Transcript Behavioral Drivers of Experience Results

The Actuarial Society of Hong Kong
Behavioral Drivers of
Mortality Experience
Texas-Wide Underwriters Conference
March 20, 2011
Tim Rozar FSA, MAAA, CERA
Vice President and Head of Global R&D
RGA Reinsurance Company
How does behavior impact mortality?
2
How does behavior impact mortality?
1) Direct Effects (Moral Hazard):
Policyholder behavior causes direct change to their own individual
mortality risk

Suicide

Lifestyle Factors (obesity, narcotics, tobacco, alcohol, driving, hobbies, etc.)
2) Indirect Effects (Anti Selection):
Policyholder behavior causes change the relative risk of the insured
pool

Applicant/agent pre-issue adverse selection

Anti-selective lapsation
3
Behavior and Mortality
Direct Effects
4
Moral Hazard
When the actions of market participants on one side are unfavorable to
the other due to misaligned incentives.
Reprinted with permission. © The New Yorker Collection from cartoonbank.com. All Rights Reserved.
5
Recent US Suicide Trend
Noticeable up-tick in suicides corresponding with Global Financial
Crisis (consistent with RGA and anecdotal industry experience)
Intentional Self Harm (Suicide)
12
11.5
11
10.5
10
2009
2008
2007
2006
2005
2004
2003
2002
2001
2000
9.5
1999
Age-Adjusted Mortality
Rate per 100,000
•
Source: CDC/NCHS
6
Suicides and Recessions
Recent research from U.S. Center for Disease Control
confirms link between economic downturns and suicides:
 “The overall suicide rate generally rose in recessions like the
Great Depression (1929-1933), the end of the New Deal (19371938), the Oil Crisis (1973-1975), and the Double-Dip Recession
(1980-1982) and fell in expansions like the WWII period (19391945) and the longest expansion period (1991-2001) in which the
economy experienced fast growth and low unemployment.
 The largest increase in the overall suicide rate occurred in the
Great Depression (1929-1933)—it surged from 18.0 in 1928 to
22.1 (all-time high) in 1932 (the last full year in the Great
Depression) - a record increase of 22.8% in any four-year period
in history. It fell to the lowest point in 2000.”
http://www.cdc.gov/media/releases/2011/p0414_suiciderates.html
7
Obesity Trends* Among U.S. Adults 1990
(*BMI ≥30, or ~ 30 lbs. overweight for 5’ 4” person)
Large volume
of research
shows link
between
mortality and
BMI
No Data
<10%
Increasing
trend in
obesity and
leads to
concerns
about mortality
rates and
trends in the
future
10%–14%
15%–19%
20%–24%
25%–29%
≥30%
8
Source, Behavioral Risk Factor Surveillance System, CDC
Obesity Trends* Among U.S. Adults 1991
(*BMI ≥30, or ~ 30 lbs. overweight for 5’ 4” person)
No Data
<10%
10%–14%
15%–19%
20%–24%
25%–29%
≥30%
9
Source, Behavioral Risk Factor Surveillance System, CDC
Obesity Trends* Among U.S. Adults 1992
(*BMI ≥30, or ~ 30 lbs. overweight for 5’ 4” person)
No Data
<10%
10%–14%
15%–19%
20%–24%
25%–29%
≥30%
10
Source, Behavioral Risk Factor Surveillance System, CDC
Obesity Trends* Among U.S. Adults 1993
(*BMI ≥30, or ~ 30 lbs. overweight for 5’ 4” person)
No Data
<10%
10%–14%
15%–19%
20%–24%
25%–29%
≥30%
11
Source, Behavioral Risk Factor Surveillance System, CDC
Obesity Trends* Among U.S. Adults 1994
(*BMI ≥30, or ~ 30 lbs. overweight for 5’ 4” person)
No Data
<10%
10%–14%
15%–19%
20%–24%
25%–29%
≥30%
12
Source, Behavioral Risk Factor Surveillance System, CDC
Obesity Trends* Among U.S. Adults 1995
(*BMI ≥30, or ~ 30 lbs. overweight for 5’ 4” person)
No Data
<10%
10%–14%
15%–19%
20%–24%
25%–29%
≥30%
13
Source, Behavioral Risk Factor Surveillance System, CDC
Obesity Trends* Among U.S. Adults 1996
(*BMI ≥30, or ~ 30 lbs. overweight for 5’ 4” person)
No Data
<10%
10%–14%
15%–19%
20%–24%
25%–29%
≥30%
14
Source, Behavioral Risk Factor Surveillance System, CDC
Obesity Trends* Among U.S. Adults 1997
(*BMI ≥30, or ~ 30 lbs. overweight for 5’ 4” person)
No Data
<10%
10%–14%
15%–19%
20%–24%
25%–29%
≥30%
15
Source, Behavioral Risk Factor Surveillance System, CDC
Obesity Trends* Among U.S. Adults 1998
(*BMI ≥30, or ~ 30 lbs. overweight for 5’ 4” person)
No Data
<10%
10%–14%
15%–19%
20%–24%
25%–29%
≥30%
16
Source, Behavioral Risk Factor Surveillance System, CDC
Obesity Trends* Among U.S. Adults 1999
(*BMI ≥30, or ~ 30 lbs. overweight for 5’ 4” person)
No Data
<10%
10%–14%
15%–19%
20%–24%
25%–29%
≥30%
17
Source, Behavioral Risk Factor Surveillance System, CDC
Obesity Trends* Among U.S. Adults 2000
(*BMI ≥30, or ~ 30 lbs. overweight for 5’ 4” person)
No Data
<10%
10%–14%
15%–19%
20%–24%
25%–29%
≥30%
18
Source, Behavioral Risk Factor Surveillance System, CDC
Obesity Trends* Among U.S. Adults 2001
(*BMI ≥30, or ~ 30 lbs. overweight for 5’ 4” person)
No Data
<10%
10%–14%
15%–19%
20%–24%
25%–29%
≥30%
19
Source, Behavioral Risk Factor Surveillance System, CDC
Obesity Trends* Among U.S. Adults 2002
(*BMI ≥30, or ~ 30 lbs. overweight for 5’ 4” person)
No Data
<10%
10%–14%
15%–19%
20%–24%
25%–29%
≥30%
20
Source, Behavioral Risk Factor Surveillance System, CDC
Obesity Trends* Among U.S. Adults 2003
(*BMI ≥30, or ~ 30 lbs. overweight for 5’ 4” person)
No Data
<10%
10%–14%
15%–19%
20%–24%
25%–29%
≥30%
21
Source, Behavioral Risk Factor Surveillance System, CDC
Obesity Trends* Among U.S. Adults 2004
(*BMI ≥30, or ~ 30 lbs. overweight for 5’ 4” person)
No Data
<10%
10%–14%
15%–19%
20%–24%
25%–29%
≥30%
22
Source, Behavioral Risk Factor Surveillance System, CDC
Obesity Trends* Among U.S. Adults 2005
(*BMI ≥30, or ~ 30 lbs. overweight for 5’ 4” person)
No Data
<10%
10%–14%
15%–19%
20%–24%
25%–29%
≥30%
23
Source, Behavioral Risk Factor Surveillance System, CDC
Obesity Trends* Among U.S. Adults 2006
(*BMI ≥30, or ~ 30 lbs. overweight for 5’ 4” person)
No Data
<10%
10%–14%
15%–19%
20%–24%
25%–29%
≥30%
24
Source, Behavioral Risk Factor Surveillance System, CDC
Obesity Trends* Among U.S. Adults 2007
(*BMI ≥30, or ~ 30 lbs. overweight for 5’ 4” person)
No Data
<10%
10%–14%
15%–19%
20%–24%
25%–29%
≥30%
25
Source, Behavioral Risk Factor Surveillance System, CDC
Obesity Trends* Among U.S. Adults 2008
(*BMI ≥30, or ~ 30 lbs. overweight for 5’ 4” person)
No Data
<10%
10%–14%
15%–19%
20%–24%
25%–29%
≥30%
26
Source, Behavioral Risk Factor Surveillance System, CDC
Obesity Trends* Among U.S. Adults 2009
(*BMI ≥30, or ~ 30 lbs. overweight for 5’ 4” person)
No Data
<10%
10%–14%
15%–19%
20%–24%
25%–29%
≥30%
27
Source, Behavioral Risk Factor Surveillance System, CDC
Obesity Trends* Among U.S. Adults 2010
(*BMI ≥30, or ~ 30 lbs. overweight for 5’ 4” person)
No Data
<10%
10%–14%
15%–19%
20%–24%
25%–29%
≥30%
28
Source, Behavioral Risk Factor Surveillance System, CDC
Obesity Trends* Among U.S. Adults 2010
(*BMI ≥30, or ~ 30 lbs. overweight for 5’ 4” person)
No Data
<10%
10%–14%
15%–19%
20%–24%
25%–29%
≥30%
29
Source, Behavioral Risk Factor Surveillance System, CDC
Cigarette Smoking
2008 VBT Male Mortality, Duration 5
1,000
350%
300%
250%
200%
10
150%
SM/NS Ratio
1000 qx (log Scale)
100
100%
1
50%
0
0%
25 28 31 34 37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 100
Attained Age at Duration 5
NS
SM
SM/NS (Right axis)
30
Smoking Trends
Pecentage Current Smokers
 Smoking prevalence in US has
dropped dramatically in past 50
years.
 Evidence of leveling-off in past
few years
 Much of the observed mortality
improvement in past 50 years
is attributable to changing mix
of smokers and non-smokers.
Percentage of Adults Who
Were Current Smokers
60
1965-2008
50
Male
40
Female
30
20
10
Source: American Lung Association
http://www.lungusa.org/finding-cures/our-research/trend-reports/Tobacco-Trend-Report.pdf
2005
2000
1995
1990
1985
1980
1975
1970
1965
0
31
Drinking and Driving
• In 2009, 10,839 people were killed in alcohol-impaired driving
crashes, accounting for nearly one-third (32%) of all traffic-related
deaths in the United States
• One in three people will be involved in an alcohol-related crash in
their lifetime.
• Every minute, one person is injured from an alcohol-related crash.
• Car crashes are the leading cause of death for teens and one out
of three of those is alcohol related.
• Teen alcohol use kills about 6,000 people each year, more than
all illegal drugs combined.
• Drivers with a BAC of 0.08% or higher involved in fatal crashes
were eight times more likely to have a prior conviction for DWI
than were drivers with no alcohol in their system
http://www.madd.org/statistics/
http://www.cdc.gov/motorvehiclesafety/impaired_driving/impaired-drv_factsheet.html
32
Cocaine and Mortality
• CRL study looked at
insurance applicants who
tested positive for cocaine.
• Prevalence rates very low,
but significantly adverse
mortality for those that test
positive.
Source: Clinical Reference Labs, “Mortality Associated with Positive Cocaine Tests”
http://www.crlcorp.com/insurance/documents/otrcocainemortalityinapplicants2010_000.pdf
33
Behavior and Mortality
Indirect Effects
34
Indirect Effects
Information Asymmetry
When market participants on one
side of a transaction have access
to better information than the other
side
Adverse/Anti Selection
When the attributes of market
participants on one side are
unfavorable to the other side due to
an asymmetric information
advantage.
35
Reprinted with permission. © The New Yorker Collection from cartoonbank.com. All Rights Reserved.
QUESTION:
How is insurance like the market
for used cars?
ANSWER:
Asymmetric information
between buyers & sellers!
36
A Simple Experiment
 You are in the market for a good quality used car and are willing to
pay up to $10,000 (USD).
37
A Simple Experiment
Good news! I have a car that I’m willing to sell for $9,000 (USD).
 It is in really good condition – trust me!
38
Assume the following market for used cars:
 75% of all used cars are in good
working order and are worth
$10,000 (USD)
 25% of all used cars are “lemons”
and are worth $2,000 (USD).
 There is no way to tell a good car
from a lemon.
39
The Market for “Lemons”
So what happens next?
 Without any verifiable information about my car or
my personal trustworthiness, you have to factor in
the risk of getting a lemon
Therefore, your expected value of my car =
(0.75*$10,000) + (0.25*$2,000) = $8,000
• You aren’t willing to spend more than $8,000, but
I won’t sell for less than $9,000 (unless I know
my car is a lemon…)
40
The Market for “Lemons”
George Akerlof’s “Lemons” model (Nobel Prize, 2001)
predicts the break-down in markets with asymmetric
information.
 The basic problem: Buyers and sellers often don’t
have access to the same information (or they can’t
verify the accuracy of the information provided by
the other party).
 Rational buyers are worried that they might be
buying a lemon, so sellers of good cars can’t get fair
value.
 This creates an unraveling market on both sides:
 Sellers with perfectly good cars can’t sell them for a fair price
 Buyers looking for good cars are increasingly likely to get
stuck with a lemon.
Nobelprize.org. 24 Mar 2011
http://nobelprize.org/nobel_prizes/economics/laureates/2001/
41
Information Asymmetry and Insurance
Applicant
 Knows detailed information
about her medical history
 Voluntarily enters insurance
market
 Demand is correlated to
riskiness
Insurer
• Has access to less
information than applicant
• Must determine riskappropriate rate for all
applicants
42
Goals of the Insurance Underwriting Process
Primary Goals
Secondary Goals
• Minimize adverse selection by
reducing information asymmetry
• Make decisions as quickly as
possible
• Accurately assess risk profile
• Minimize intrusiveness to applicant
and agent
• Uncover existence and severity of
medical impairments
• Provide sentinel to discourage
agent/applicant anti-selection
• Minimize underwriting costs
• Maximize case placement rates
43
Underwriting reduces the information asymmetry
between applicants and insurer
Applicants
U/W
Filter
Declines
Insureds
44
“Simplified Issue” improves secondary u/w goals
but a few “lemons” may get through
Full U/W
Filter
SI U/W
Filter
45
Simplified Issue Experience
 In U.S., industry experience for simplified issue business is much
worse than for underwritten policies sold at similar face amounts.
Why?
Placed case mortality is determined by:
A)
Applicant Pool Mortality
PLUS
B)
Underwriting Filter
46
What happens to the fully underwritten declines?
Full U/W
Filter
SI U/W
Filter
47
What happens to the Fully UW Declines?
48
The applicant pool begins to change when Fully U/W
declines become SI applicants
Full U/W
Filter
SI U/W
Filter
49
What About Those Who Don’t Bother Applying for Fully UW?
50
Reduced sentinels may encourage more
adverse changes to applicant pool
Full U/W
Filter
SI U/W
Filter
51
Anti-selective lapsation can lead to
additional deterioration of mortality
Full U/W
Filter
SI U/W
Filter
52
The market can unravel as information
asymmetry and anti-selection increase.
Full U/W
Filter
SI U/W
Filter
53
Post-Level Term Experience
Post-level term experience is one of the clearest observable
demonstrations of anti-selective policyholder behavior.
Annual Premium
Sample Premiums
Male Age 45 Super Preferred NS
10 Year Term, $500,000
$7,845
Duration 11
$330 Level Period
Premium
1
2
3
4
5
6
7
Level Period
8
9 10 11 12 13 14 15 16 17 18 19 20
Tail Period
54
Post-Level Term Experience
• Sharp increase in premium
after level period leads to
large anti-selective shock
lapse.
Lapse Rate
70%
60%
50%
40%
30%
20%
10%
0%
6
7
8
9
10
11
Level Period
13+
Tail Period
2008 VBT Mortality Ratio
300%
• Mortality on persisting
policyholders is substantially
worse in the post-level
period.
12
250%
200%
150%
100%
50%
0%
6
7
8
9
Level Period
10
11
12
13+
Tail Period
55
Post-Level Term Experience
• Strong correlation between the size of a company’s shock lapse and
the amount of post-level period mortality deterioration – the larger the
shock lapse, the worse the post-level period mortality.
500%
Mortality Relative
to Durations 6-10
450%
400%
350%
300%
250%
200%
150%
100%
50%
0%
0%
20%
40%
60%
Duration 10 Shock Lapse
80%
100%
56
Impact of Genetic Testing on
Insurance Purchasing Behavior
•
A genetic test exists for ApoE (e4) and other genetic markers of
increased risk for Alzheimer’s disease
•
REVEAL Study: Randomized controlled trial to evaluate impact of
genetic education on adult children of Alzheimer’s Disease (AD)
patients
Control group - Told of AD risk based on age, gender, family history
Intervention group - Told of AD risk based on age, gender, family history
and ApoE genotype
•
Overall, e4 positive subjects 5.8 times more likely to increase LTCI
coverage than those who did not receive ApoE genotype disclosure
Zick, CD et al. Genetic Testing for Alzheimer’s disease and its Impact on Insurance
Purchasing Behavior. Health Affairs 2005 (March); 24:483-90.
57
Large Face Term Mortality
• Intuition suggests large face amount policies should have better
mortality than any other policies:
• Higher socio-economic class
• More rigorous underwriting requirements
However, U.S. industry experience beginning to suggest mortality
is actually worse at higher face amounts.
1) Anti-selection: An applicant’s demand for insurance is positively
correlated with their risk
2) Moral Hazard: Higher suicide and other accident mortality at
larger face amounts
58
Solutions
Three opportunities to mitigate or manage the impact of behavior on
mortality:
1. Pre-issue (applicant pool)
2. Underwriting
3. Inforce management
59
Applicant Pool
• Improve sentinels
• Broaden exposure base
•
Higher participation rates lead to reduced anti-selection ( e.g. COLI, car insurance,
non-contributory group coverage, single payer systems)
• Link insurance sale to need or life event
•
Financial planning, education savings, home mortgage
• Price competitively
•
Don’t discourage good risks from applying (price increases can lead to death spiral)
• Target marketing/Pre-Screening
• Incentives to encourage applicant “signaling”
•
e.g. Progressive “Snapshot” program
60
Underwriting Filter
• Maintain sound underwriting practices
•
•
•
Don’t forget about “primary” underwriting goals
Gather the evidence required to assess risk appropriately
Reflexive interviews may bring more clarity to application disclosures
• Improve vigilance on financial underwriting
•
Coverage amount should be proportional to need, not risk
• Increase insurers’ access to verifiable information on applicants to
reduce information asymmetry
•
Health and prescription drug histories, prior underwriting disclosures, motor vehicle
records, criminal history, cognitive screening, etc.
61
Inforce Management
• Maintain sound claims management practices
• Enact “smart” policyholder retention/conversion programs
• Avoid abnormally rich benefits or policy wording that may encourage
moral hazard (or malingering).
• Identify targeted cross-marketing opportunities
• Encourage favorable policyholder behavior
• Wellness credits for health maintenance (e.g. Discovery Vitality in South
Africa)
62
Conclusions
 Behavioral dynamics should play a big role in how actuaries and
underwriters think about setting mortality expectations.
 Changes in general population lifestyle factors could have a profound
impact on forward-looking mortality expectations
 Sound underwriting will focus on analyzing all reasonable information
to identify applicant behaviors that could impact mortality risk
 Do not ignore the “lemons” problem created by increased information
asymmetry in simplified issue products.
 Product design should carefully consider the potential for moral
hazard and anti-selection
63
The Actuarial Society of Hong Kong
Behavioral Drivers of
Mortality Experience
Texas-Wide Underwriters Conference
March 20, 2011
Tim Rozar FSA, MAAA, CERA
Vice President and Head of Global R&D
RGA Reinsurance Company